Last interglacial sea levels within the Gulf of Mexico and northwestern Caribbean Sea

During the last interglacial (LIG) the volume of additional water in the world’s oceans was large enough to raise global sea levels about 6–9 m higher than present levels. However, LIG sea levels vary regionally and those regional differences hold clues about the past distribution of ice sheets and local rates of subsidence and tectonic uplift. In this study, I used a standardized database template to review and summarize the existing constraints on LIG sea levels across the northern Gulf of Mexico and Caribbean shoreline of the Yucatán Peninsula. In total, I extracted 32 sea-level indicators including the insertion of 16 U-series ages on corals, 1 electron spin resonance age, 2 amino acid racemization ages, and 26 luminescence ages. Most dated sea-level indicators for the northern Gulf of Mexico are based on optically stimulated luminescence (OSL) ages of beach deposits of a mappable LIG shoreline. This shoreline extends from the Florida Panhandle through south Texas but is buried or removed by the Mississippi River across most of Louisiana. A similar feature is observed in satellite images south of the Rio Grande within the Mexican portions of the Gulf of Mexico but has yet to be dated. Elevations measured on portions of this feature close to the modern coast point to sea levels less than 1 m to ∼ 5 m higher than present for much of the northern Gulf of Mexico. However, a few, albeit undated, portions of the same shoreline located at more inland locations point to sea levels up to +7.2 m, attesting to up to 7 m of differential subsidence between the inland and coastal sites. Across the Yucatán Peninsula, U-series dating of corals has provided the main index points for LIG sea levels. Other carbonate coastal features such as beach ridges and eolianites have also been described but rely on corals for their dating. The maximum elevation of the LIG coral-based relative sea-level (RSL) estimates decrease from around +6 m across the Caribbean shoreline of the Yucatán Peninsula near Cancún, Mexico, to as low as −6 m to the south beneath the southern atolls of Belize, although discussion continues as to the validity of the ages for these southern corals. If these lower-elevation corals are LIG in age, their below-present elevations may be a result of vertical motion along faults dipping into the Cayman Trough. South of Belize only one purported LIG coral has been dated on the Isla de Roatán off the coast of Honduras at a likely tectonically uplifted elevation of 37.2 m. Thus the elevation of LIG sea levels within the inland siliciclastic shorelines of Guatemala and Honduras as well as the southwestern Gulf of Mexico remains poorly constrained and a potential venue for future research. The database described in this paper is available open access in spreadsheet format as Simms (2020), at this link: https://doi.org/10.5281/zenodo.4556163. Published by Copernicus Publications. 1420 A. R. Simms: Last interglacial sea levels within the Gulf of Mexico and northwestern Caribbean Sea

1 Introduction ( Fig. 1), with its eastern portions still debated as to their origin (Otvos, 1997). In many locations, it still maintains the ridge and swell topography of old beach ridges (Fig. 2). Simms et al. (2013) obtained six OSL ages from four cores up to ~5 m in length from two portions of the barrier. Three of the ages came from the core of the barrier, while three additional ages were 130 obtained from the reworked top veneer of the feature. The three ages from the core of the feature were 119.0±7.0 ka (WALIS LUM_ID #145), 120.1±8.40 ka (WALIS LUM ID #119) and 127.9±8.70 ka (WALIS RSL ID #118) (Simms et al., 2013). The younger ages ranging between 1.3 and 57.0 ka and all coincided with time periods of known heightened aeolian activity (Otvos, 2004;Simms et al., 2013).

135
The elevation of the LIG shoreline varies across its expression in Texas. Paine (1993) provided one of the first attempts at a rigorous quantitative estimate of RSL change at the last interglacial based on the Ingleside shoreline in Texas. In order to quantify subsidence across the Gulf of Mexico at different time scales, Paine (1993) noted the maximum elevation of shell horizons in borings was 2-m above modern sea level (general definition) with the highest in situ oysters (Crassostrea virginica) at an elevation of 0.5 m. The Ingleside shoreline attains higher elevations but a portion of that elevation is late Pleistocene 140 coastal dunes that covered the shoreline at the time of its formation (Wilkinson et al., 1975). Paine (1993) subtracted the 2-m of elevation from a "global" 8-m sea level highstand at the LIG to suggest a long-term subsidence rate of 0.05 mm/yr for the central Texas Coast. Although the exact facies represented by the shell horizon is poorly constrained with respect to past tidal datums, as it comes from boring descriptions (Paine, 1993), it likely represents deposition within a foreshore or barrier flat environment, which do not extend to elevations of more than ~1 m above modern sea levels along the Texas Coast (Brown et 145 al., 1976;Rodriguez et al., 2004;Simms et al., 2006). The shell deposits may have originated in deeper water. Although sandy lithosomes (e.g upper-shoreface sands) along the modern central Texas coast extend into water depths as great as 12 m (Rodriguez et al., 2001), they likely do not represent deposition within water depths any greater than 2.5 m as water depths greater than that would place coeval Ingleside eolian deposits underwater (Wilkinson et al., 1975). Assuming the Ingleside eolian cap was originally eolian in origin and simply reworked by later dry phases of climate (Otvos, 2004;Simms et al., 150 2013), I assign a LIG sea-level range of +1 m to +4.5 m for the shell horizons mentioned by Paine (1993)(WALIS RSL ID #915). Simms et al. (2013) took a different approach to estimate paleo-RSL from the Ingleside by mapping the feature in an geographical information system (GIS) software package using soil survey maps and determining its elevation from the United 155 States Geological Survey's (USGS) National Elevation Dataset (NAD) digital elevation model (DEM). Assuming the Ingleside was a LIG barrier island (Price, 1933;Paine, 1993) similar to the modern barrier islands of the Texas Gulf Coast, Simms et al. (2013) subtracted the closest modern equivalent barrier island elevation from the elevation of each of the Ingleside shoreline segments of the Texas Coast. Assuming the preserved Ingleside segments represent the highest RSLs reached during the LIG, the resulting calculations lead to a range of RSL differences at the LIG across the Texas coast from a high of 7.2 m 160 for the Vidor segment (WALIS RSL ID #778) to a low of 0.2 m for the Hoskins Island segment (WALIS RSL ID #774) (Fig. https://doi.org/10.5194/essd-2020-253 Open Access Earth System Science Data Discussions Preprint. Discussion started: 28 September 2020 c Author(s) 2020. CC BY 4.0 License.
3). However, the Vidor segment (Orange of Otvos, 1997) and Hoskin Islands Segments may not represent a barrier (Otvos, 1997) and have yet to be dated. If the Vidor segment represents a different age or depositional environment, then the highest non-contested LIG barrier in East Texas would be the Fannett sector, which has well-preserved beach ridge features, with an elevation of +5.8 m (WALIS RSL ID #777) (Otvos, 1997;Simms et al., 2013). Regardless, variability in the elevations of the 165 LIG shorelines is more than 5 m. This variability was interpreted by Simms et al. (2013) to represent differential subsidence across the Texas coastal plain. Simms et al. (2013) found a RSL difference of 2.8 m for the same segment of the Ingleside (Live Oak; WALIS RSL ID #772) that Paine (1993) found a shell horizon at an elevation of 2 m (Willow Creek; WALIS RSL ID #915).

Northeastern USA Gulf of Mexico 170
The Gulfport Shoreline, in some locations also known as the Palmico Shoreline, has a similar expression and elevation as the Ingleside shoreline of Texas (Otvos, 1972). It rises between +5 m and +9 m and can be traced from the Pearl River, Mississippi to near the Apalachicola delta of the Florida Panhandle (Otvos, 2005). It has been dated in three general locations using OSL.

185
The original publications of Blum et al. (2003) and Otvos (2005) provide little information about the elevations of the barrier segments dated. However, Rodriguez and Meyer (2006)

195
As many of the LIG sites along the northeastern Gulf of Mexico lack quantitative estimates of the elevation of RSL at the LIG, we followed the methods of Simms et al. (2013) to assign a RSL elevation for the LIG. This estimate was determined by subtracting the average elevations of the closest modern barrier islands from the average elevations of the five segments of the Gulfport Shoreline dated ( Fig. 4; Table 1). Mapping the margins of the LIG shoreline along the northeastern Gulf of Mexico using soil surveys is not as straight forward as it is along the Texas coast due to the sandier nature of much of the northeastern 200 Gulf of Mexico coastal plain and shelf (the Gulfport Shoreline is not bordered along its inland margins by a muddy unit as the Ingleside Shoreline is in Texas). From this approach I obtained RSL estimates at the LIG for the English Lookout,  Rodriguez and Meyer (2006). In each case, I deferred to the elevations derived from the GPR rather than the average elevations of the shoreline segments within the WALIS 210 database.

Mexican Gulf of Mexico
A feature similar to the Ingleside appears to continue south along the Gulf of Mexico south of the USA/Mexico border to Soto la Marina, Tamaulipas (Price, 1958) and possibly farther south into Veracruz-Llave but has yet to be dated (Wilhelm and Ewing, 1972;Hernandez-Santana et al., 2016;Figs. 1 and 2). Near Soto la Marina these features are dotted with small ponds 215 similar to the blow-out features common to the Ingleside barrier across Texas (Price, 1933;Otvos, 2005;Simms et al., 2013).
However, their LIG age has not been verified and thus no data for these features has been input into the WALIS database.
More work mapping and dating this potential LIG shoreline is warranted.

Yucatan Peninsula
Dated LIG beach ridges and reefs have been identified and studied across many locations of the Yucatan coastlines of Mexico 220 and neighboring Belize (Fig. 5). Additional constraints on LIG sea levels from the Yucatan have been reported based on speleothems within caves near the Mexican LIG beach ridges and coral reefs. These are the subject of a separate compilation within WALIS but are briefly discussed with reference to the other data reviewed in this study.

Mexico
A prominent set of LIG calcarenite beach ridges extends across much of the northeastern portion of the Yucatan (Szabo et al., 225 1978;Ward and Brady, 1979). The calcarenite beach ridge plain extends 150 km from Cancun to Xel Ha with a width of 0.5 to 4 km and thicknesses ranging from 3 to 10 m (Ward and Brady, 1979) (Fig. 5). The strandplain is underlain by a caliche developed over older Pleistocene coral-bearing limestones (Fig. 6). In addition, a few isolated Diploria and Montastrea corals in growth position overlie the caliche crust but are covered by the overlying calcarenite beach deposts (Ward and Brady, 1979;Szabo et al., 1978). Szabo et al. (1978) obtained 5 U-series ages on corals reworked into the calcarenite beach ridge deposits 230 as well as the underlying isolated in situ corals. Three of the corals reworked into the overlying calcarenite beach deposits dated to 121±6 ka, 123±6 ka, and 120±6 ka (WALIS RSL ID #438). In situ or only lightly abraded corals of Montastrea sp.
found ontop of the caliche crust dated to 123±6 ka (WALIS RSL ID #439) and 125±15 ka (WALIS RSL ID #440; Szabo et al., 1978). In addition to the corals within and underlying the calcarenite beach ridges between Cancun and Xel Ha, Szabo et al. (1978) also obtained an age from a reef coral on the nearby island of Cozumel that returned an age of 121±6 ka (WALIS 235 RSL ID #441).
Although the calcarenite beach ridges reached elevations of 10 m (Szabo et al., 1978), they are capped by an eolianite facies ( Fig. 6). The base of the calcareous beach facies with cross-bedding lies at elevations of +3.5 m to +6.5 m above present sea level (general definition; Szabo et al., 1978). A more detailed stratigraphic description of the deposits by Ward and Brady 240 (1979) suggests the boundary between the upper shoreface and foreshore/backshore is found at ~+4.8 m across the calcarenite strandplain. Based on the difference in elevation between that LIG contact and the modern upper shoreface/forshore contact (contact elevation not given), Ward and Brady (1979) argue that RSL during the LIG was between +5 m and +6 m. I therefore assign a value of 5.5±1.5 m to the calcareous beach-ridge derived LIG sea-level estimate (WALIS RSL ID #438). The extra 1m of error was added to account for the uncertainties associated with how the elevation was measured, what datum was used, 245 and the details of the modern analogue. The two lower elevation Montastrea corals, which inhabit water depths of -1.1 m to -17 m (Lightley et al., 1982;Hibbert et al., 2016) at +2 m as well as a third on the nearby island of Cozmel also at +2 m are consistent with such a RSL assignment. Alternatively, accounting for the modern elevation distribution of Montastrea of -9.7 +8.6/-7.3 m (Hibbert et al., 2016) results in an RSL estimate of 11.7 +8.6/-7.3 m for these three samples (WALIS RSL IDs #439, #440, #441), a much higher elevation but larger error range than the original interpretations of Szabo et al. (1978). An 250 earlier higher sea level is not out of the realm of possibility as the overlying calcareous strandplain has a regressive character (Ward and Brady, 1979) but an equivalent shoreline at 5+ m higher than the calcarenite strandplain has yet to be identified. In addition, it would conflict with other data from very-well preserved reefs in the same area studied by Blanchon et al. (2009).
Therefore within the WALIS database, I defer to the original interpretations of Szabo et al. (1978) and Ward and Brady (1979) of a RSL meaning for these three corals equivalent to their overlying beach ridges of 5.5±1.5 and include them in the RSL 255 https://doi.org/10.5194/essd-2020-253
These corals have been dated using U-series ages by Gischler et al. (2000) and Mazzullo (2006). In addition, Mazzullo (2006) obtained two additional amino acid racemization ages from the corals. annularis coral dredged from 2.3 m depth that dated to 165.5±1.1 ka, but was deemed unreliable given its high Th content.
Using the Caribbean distribution of these two species by Hibbert et al. (2016) places LIG RSLs at +1.8 +1.2/-7.9 m (WALIS 315 U-Series ID #6) and +10.2 +8.6/-7.3 m (WALIS RSL ID #448), respectively. The much higher estimate of the former is largely based on the habitat range of M. annularis, which is much deeper than A. palmata. However, the M. annularis was found in association with A. palmata (Mazzullo, 2006) and likely represents shallower water -closer to the lower limit of the depth range quoted. Two specimens of the gastropod Strombus gigas gave AAR ages equivalent to the LIG from an elevation of +1.2 m approximately 7 km southwest of Reef Point (WALIS AAR ID #s 129,130;Mazzullo, 2006). The gastropod 320 inhabits very shallow waters but can be found in water depths as great as 60 m (Randall, 1964), and thus only confirms the RSL elevation limits placed by the corals but does support the age assignment of the reefs. However, Gischler et al. (2000) mentions that the facies the coral were obtained from also includes specimens of the coral A. palmata, which would suggest lower sea levels on the order of -7.5 +1.2/-7.9 m (WALIS U-Series ID #8) and -6.5 +1.2/-7.9 represent the LIG highstand in Belize with the subsequent lower elevations to the south and east along the Belize margin a reflection of neotectonic activity or deposition at a time younger than the LIG.

Honduras
Only a handful of possible LIG deposits have been located in Honduras. Late Pleistocene limestones with in situ specimens of Montastrea sp. and Acropora cervicornis have also been reported from the Swan Islands (Ivey et al., 1980;Fig. 8) but have yet to be dated. They reach elevations up to 14 m above modern sea level.
Both regions are likely heavily influenced by tectonic activity due to their development and growth across uplifted tectonic blocks along the Montagu/Swan Islands fault system (Cox et al., 2008). The mainland coast of Honduras is a well-developed 365 siliciclastic coastline with prevalent presumably Holocene beach ridges but no LIG shorelines have been mapped across it to date.

Datums
With the exception of the new work in this study and the works of Burdette et al.

Elevation measurements
The rest of the studies defined mean sea-level according to the generic definition and provided little detail as to how the 380 only been preliminarily dated (e.g. Otvos, 2005) and more works is required to nail down their ages and relationship to former sea levels.
Older Pleistocene reefal units are present across the Yucatan Peninsula (e.g. Ward and Brady, 1979;Ferro et al., 1999;Gischler et al., 2010) but have not been well dated nor been used to constrain the elevations of pre-LIG highstands. Speleothems that 420 may help constrain older sea levels dating as far back as MIS11 have been identified within Quintana Roo (Steidle et al., 2020).
Those results have yet to be published outside of meeting abstracts, but are likely forthcoming.

Holocene sea-level indicators
Middle-to-late Holocene sea levels are well constrained in the region with several site-specific reconstructions as well as compilations available for the northern Gulf of Mexico (Tornqvist et al., 2004;Simms et al., 2007;Milliken et al., 2008;Livsey 425 and Simms, 2013) as well as the Caribbean (Toscano and Macintyre, 2003;Gischler and Hudson, 2004;Khan et al., 2017).
The records become sparser for the early Holocene and late glacial periods. One discussion that has repeatedly resurfaced within the northern Gulf of Mexico is the possibility of a mid-Holocene highstand (e.g. Tanner et al., 1989;Blum et al., 2002) but currently appears to have fallen out of favor (Otvos, 2001;Simms et al., 2009).

Uncertainty and data quality 430
The amount of uncertainty in the age and elevation of the LIG sea-level indicators varies by location. The shoreline along the northern Gulf of Mexico is likely LIG in age but very few of the existing ages have the accuracy or precision to determine when within the generally accepted 115-129 ka time period it formed. The average error of the 24 OSL measurements thought to have been derived solely from LIG deposits is 10.4 ka, far too large to determine when within the LIG the feature(s) formed.
Because few of the studies on the LIG shoreline to date have included detailed facies descriptions of the shoreline deposits, 435 the elevations are probably accurate to within 2-3 m of the former highstand elevation. In addition, the lack of estimates of subsidence independent of the LIG elevation at each site also contributions to the uncertainty of LIG RSLs along the Gulf of Mexico. This uncertainty due to subsidence is likely on the order of <5 m (Paine, 1993;Simms et al., 2013), but these estimates are in need of analyses independent of the LIG shoreline elevations.

440
The data from the northeastern Yucatan Peninsula probably provides the best estimates of RSL during the LIG for the region surveyed in this study. The analysis of Blanchon et al. (2009) includes the most detailed facies analysis of coral reef deposits within the region leaving LIG RSL elevation estimates to within 1 m. In addition, their screened U-Th ages appear to be able to distinguish early from late LIG times. The earlier study of the Mexican Yucatan Peninsula by Szabo et al. (1978)